Title: Improving history match using artificial neural networks
Authors: Mohamed A. Habib; Abdulaziz M. Abdulaziz; Abdel-Sattar A. Dahab
Addresses: The General Petroleum Company, 8 Dr. Mostafa Abu Zahra St., Nasr City, Cairo, Egypt ' Mining, Petroleum, and Metallurgical Engineering Department, Faculty of Engineering, Cairo University, Giza, Egypt ' Mining, Petroleum, and Metallurgical Engineering Department, Faculty of Engineering, Cairo University, Giza, Egypt
Abstract: Of late, the application of the artificial intelligence in oil industry has been increasing and seems promising in reservoir simulation studies. In this research, an artificial neural network (ANN) was attempted to achieve a preliminary history match with fewer simulation runs for two different reservoir simulators; compositional and black oil. In the compositional model, local operations are utilised to control fluids flow through baffle system. With only 15 runs, a very good match is achieved in all wells for both drainage reservoir pressure and the gas oil ratio, but water cut required a manual tuning. In the black oil model, a fluvial system is constructed with six oil producers and four water injectors. The results showed an obvious improvement to the trend of bottom hole pressure and water cut generated from eight runs compared to the base case. Such results encourage adopting artificial intelligence techniques towards automated history match.
Keywords: history match; reservoir simulation; artificial intelligence; neural networks; pressure; water cut; gas oil ratio; GOR; permeability.
International Journal of Petroleum Engineering, 2016 Vol.2 No.4, pp.302 - 319
Received: 28 Oct 2016
Accepted: 12 Feb 2017
Published online: 12 May 2017 *